Metrics Server

Metrics Server in Thalassa Cloud Kubernetes

The Metrics Server is a vital component in Thalassa Cloud Kubernetes, providing resource usage data for CPU and memory consumption of running Pods and Nodes. It is primarily used by the Horizontal Pod Autoscaler (HPA) and the kubectl top command to enable real-time monitoring and scaling decisions.

Thalassa Cloud pre-installs and fully manages Metrics Server, ensuring seamless access to resource metrics without requiring additional configuration.

How Metrics Server Works

Metrics Server collects resource usage statistics from the Kubernetes kubelet API on each node. It then aggregates and exposes this data through the Kubernetes Metrics API, making it available for tools and controllers that require real-time resource monitoring.

Key Functions of Metrics Server:

Provides CPU and Memory Metrics: Allows kubectl top to display resource consumption.
Enables Horizontal Pod Autoscaling (HPA): Supplies real-time metrics for automated scaling decisions.
Lightweight and Efficient: Designed to work with large clusters with minimal overhead.
Cluster-Wide Visibility: Ensures access to aggregated resource usage data for monitoring tools.

Key Features

Feature	Description
Real-Time Resource Metrics	Provides live CPU and memory usage of Pods and Nodes.
Enables Autoscaling	Supplies data for Kubernetes Horizontal Pod Autoscaler (HPA).
`kubectl top` Integration	Allows real-time monitoring with `kubectl top nodes` and `kubectl top pods`.
Lightweight and Efficient	Minimal performance overhead, suitable for production workloads.

Usage

Viewing Node Metrics

To check resource usage for nodes:

kubectl top nodes

Example output:

NAME           CPU(cores)   CPU%   MEMORY(bytes)   MEMORY%
k8s-node-1    250m         5%     1024Mi          30%
k8s-node-2    300m         6%     1100Mi          32%

Viewing Pod Metrics

To check resource usage for Pods:

kubectl top pods -n my-namespace

Example output:

NAME          CPU(cores)   MEMORY(bytes)
web-app-1     100m         256Mi
web-app-2     150m         300Mi

Checking if Metrics Server is Running

To verify that the Metrics Server is active in your cluster, run:

kubectl get deployment -n kube-system metrics-server

A healthy cluster should return an active deployment.

Troubleshooting Metrics Server

If the Metrics Server is not providing data, consider the following checks:

Ensure the Metrics API is available:
```
kubectl get apiservices | grep metrics
```

Check Metrics Server logs for errors:

kubectl logs -n kube-system deployment/metrics-server

Ensure API communication is not blocked by network policies.

Konnectivity Agent Node Problem Detector